Dear Shinichi and List, you could try to re-estimate ultrametric trees given the topology of the (rooted) non-ultrametric trees and the original molecular data.
Here is a small example using either a distance based approach or maximum likelihood (ML): # load some data library(phangorn) data(Laurasiatherian) dm <- dist.ml(Laurasiatherian) treeNJ <- midpoint(nj(dm)) # let the neighbor joining tree the one we want to make ultrametric # we can do with distance based methods (non-negative least-squares) # reusing the distance matrix we computed above treeUltra <- nnls.phylo(treeNJ, dm, rooted=TRUE) # and we can optimize the ultrametric tree using ML fit <- pml(treeUltra, Laurasiatherian, k=4) # here it is important to provide the ultrametric tree fit <- optim.pml(fit, optRooted=TRUE, model="GTR", optGamma=TRUE) # if you like you can make tree rearrangements fit2 <- optim.pml(fit, optRooted=TRUE, model="GTR", optGamma=TRUE, rearrangement = "NNI") par(mar = c(1,1,4,1)) par(mfrow = c(2, 2)) plot(treeNJ, main = "Original tree") plot(treeUltra, main = "Ultrametric (distance)") plot(fit, main = "Ultrametric (ML)") plot(fit2, main = "Ultrametric (ML + NNI)") Just a warning, the optimization of the ultrametric ML function is much slower than the non-ultrametric one (a reason why non-ultrametric optimization is so popular) and less well tested. Cheers, Klaus On Mon, Oct 31, 2016 at 9:46 PM, Giulio V. Dalla Riva <gv...@uclive.ac.nz> wrote: > My 2cents on this. > > > TL;DR: there *may* be good information in the non ultrametric tree, which > we may *not* want to throw away lightheartedly. > > Long version: A phylogenetic model variance-covariance matrix is not, in > general, the same thing of the pairwise phylogenetic distance matrix. They > are proportional only in the case the rate of phenotypic evolution is > constant across the tree (e.g., a Brownian Motion with one regime). In that > case the "time available for phenotypic evolution" is, indeed, a predictor > of the expected variance (and covariance) of a trait. But that is not > always the case. > > > And, indeed, "a model allowing for variation in rate of trait evolution" > *does* correspond to a transformation of the "original" (time calibrated, > ultrametric) phylogenetic tree into a tree, in general non ultrametric, > where the length of the branches correspond to the expected amount of > evolution. i.e., a model of punctuated evolution on the original tree > corresponds to a Brownian Motion on a rescaled tree where each (inner or > pendant) branch has the same length. > > > A timed tree from molecular data may, actually, highlight how certain > lineages evolved under a different mutation/selection regime to other > branches. And the power to detect accelerated/slowed evolutionary rate may > be higher from molecular data than from phenotypic data, so throwing that > information away forcing the tree to be ultrametric may not be the best > strategy. For example, there may be some particular trait that exhibit a > different pattern from the overall evolutionary rate of the set of species > we are looking at (e.g., a very labile trait evolving fast on some lineages > where evolution is slow, or vice versa). > > > I think there's a question there that need to be addressed: why were the > trees non ultrametric at the first place? > > > Giulio Valentino Dalla Riva > > Beaty Biodiversity Research Centre > University of British Columbia > Vancouver, Canada > ________________________________ > From: R-sig-phylo <r-sig-phylo-boun...@r-project.org> on behalf of > Alejandro Gonzalez Voyer <alejandro.gonza...@iecologia.unam.mx> > Sent: October 31, 2016 12:56:02 PM > To: Emmanuel Paradis > Cc: R-phylo Mailing-list; Shinichi Nakagawa > Subject: Re: [R-sig-phylo] Making ultrametric trees > > Hi, > > >> Note that vcv(nonultrametric_tree, cor=TRUE) is different from > >> vcv(chronoMPL(nonultrametric_tree), cor = TRUE), which is relates to my > >> previous question (which one is preferred?) > > > > It (mainly) depends whether you want to assume a "clock-like" evolution > for the traits you are studying. > > Assuming that Sinichi is working with contemporaneous species, the > assumption of comparative analyses is that the time available for > phenotypic evolution (the off diagonal element of the VCV matrix) is the > same for all species. The rate of phenotypic evolution may of course differ > among species, but I think that this is not the same as allowing the > expected change in the traits to depend on the branch lengths of a > non-ultrametric tree, as would be the case in vcv(nonultrametric_tree, > cor=TRUE). In such a case, a model allowing for variation in rate of trait > evolution in conjunction with an ultrametric tree would be more > appropriate. Of course, as pointed out by Brian, estimating divergence > times is no piece of cake and there may be uncertainty associated with that > step of phylogeny estimation as well. > > Cheers > > Alejandro > > _______________________________________________ > Dr Alejandro Gonzalez Voyer > Newton Advanced Fellow > Investigador Asociado C > > Laboratorio de Conducta Animal > Instituto de Ecología > Circuito Exterior S/N > Ciudad Universitaria > Universidad Nacional Autónoma de México > México, D.F. > 04510 > México > > Tel: +52 55 5622 9044 > E-mail: alejandro.gonza...@iecologia.unam.mx > > Lab website: www.alejandrogonzalezvoyer.com<http://www. > alejandrogonzalezvoyer.com> > > > El 31/10/2016, a las 1:30 p.m., Emmanuel Paradis < > emmanuel.para...@ird.fr> escribió: > > > > Le 30/10/2016 à 03:09, Shinichi Nakagawa a écrit : > >> Dear Emmanuel and Alejandro > >> > >> Many thanks for your replies. > >> > >> Emmanuel, I understand what you said if I use vcv(nonultrametric_tree, > >> cor=FALSE) in a phylogenetic comparative method (PCM). But if I use > >> vcv(nonultrametric_tree, cor=TRUE), isn't this the same as assuming a > >> ultrametric tree? I may be missing something here? > > > > No, this is not necessarily the same thing since the off-diagonal > elements of the matrices may be different depending on how branch lenghts > are changed by chronoMPL (or other procedures you might use to make the > tree ultrametric). > > > >> Note that vcv(nonultrametric_tree, cor=TRUE) is different from > >> vcv(chronoMPL(nonultrametric_tree), cor = TRUE), which is relates to my > >> previous question (which one is preferred?) > > > > It (mainly) depends whether you want to assume a "clock-like" evolution > for the traits you are studying. > > > > Cheers, > > > > Emmanuel > > > >> Many thanks > >> > >> Shinichi > >> > >> On Sun, Oct 30, 2016 at 9:53 AM, Alejandro Gonzlaez Voyer > >> <alejandro.gonza...@iecologia.unam.mx > >> <mailto:alejandro.gonza...@iecologia.unam.mx>> wrote: > >> > >> Hi, > >> > >> I would venture in adding that with an additive tree the differences > >> in evol rates among species are determined by the genes used to > >> estimate the tree, given differences in substitution rates among > >> genes. I do think its important to bear in mind. > >> > >> Cheers > >> > >> Alejandro > >> > >> Alejandro Gonzalez Voyer > >> Instituto de Ecologia > >> UNAM > >> > >> > >> > >> > >> > On Oct 29, 2016, at 2:45 PM, Emmanuel Paradis > >> <emmanuel.para...@ird.fr <mailto:emmanuel.para...@ird.fr>> wrote: > >> > > >> > Sure they are different. I forgot to mention that. > >> > > >> > In comparative analyses, the expected quantity of trait change (= > >> the variance in the VCV matrix) is given by the product of the rate > >> of evolution of the trait with branch length. So using a > >> non-ultrametric tree is a way to assume different rates of evolution > >> for each branch. With an ultrametric tree, you somehow assume a > >> "clock-like" trait evolution (i.e., the quantity of change from the > >> root of the tree to the tips is the same). > >> > > >> > BTW, the first command you give below should be (misplaced right > >> parenthesis): > >> > > >> > vcv(chronoMPL(nonultrametric_tree), cor = TRUE) > >> > > >> > Best, > >> > > >> > Emmanuel > >> > > >> >> Le 29/10/2016 à 13:12, Shinichi Nakagawa a écrit : > >> >> Dear Emmanuel > >> >> > >> >> Many thanks for your swift reply. > >> >> > >> >> If we get a correlation matrix, via chronoMPL [i.e. > >> >> vcv(chronoMPL(nonultrametric_tree,cor=T))], this correlation > >> matrix is > >> >> different from a correlation matrix obtained directly from vcv > [i.e. > >> >> vcv(nonultrametric_tree, cor=T)]. > >> >> > >> >> Could you enlighten me about differences and why chronoMPL method > >> might > >> >> be preferred? > >> >> > >> >> Best wishes and thanks in advance. > >> >> > >> >> Shinichi > >> >> > >> >> On Sat, Oct 29, 2016 at 8:50 PM, Emmanuel Paradis > >> >> <emmanuel.para...@ird.fr <mailto:emmanuel.para...@ird.fr> > >> <mailto:emmanuel.para...@ird.fr <mailto:emmanuel.para...@ird.fr>>> > >> wrote: > >> >> > >> >> Dear Shinichi, > >> >> > >> >> A fast solution is provided by the function chronoMPL in ape; > >> it does > >> >> not require calibration points (see the references in the help > >> page of > >> >> this function for the assumptions). The function chronos (also > >> in ape) > >> >> is another possibility but it needs at least one calibration > point > >> >> (though this can be fixed to unity at the root). > >> >> > >> >> Note that a correlation matrix can be calculated from a > >> non-ultrametric > >> >> tree: see the function vcv. > >> >> > >> >> Best, > >> >> > >> >> Emmanuel > >> >> > >> >> Le 29/10/2016 à 05:26, Shinichi Nakagawa a écrit : > >> >> > Dear List > >> >> > > >> >> > We would like make several non-ultrametic trees (based on > >> >> molecular data; > >> >> > downloaded from treeBASE) into ultrametric trees using R. > Could > >> >> you advice > >> >> > us what is the best and the most general way of doing this > >> please? > >> >> > > >> >> > We would like to use resultant ultrametric trees in > >> phyologenetic > >> >> > comparative analysis (by turning into correlation matrices). > >> >> > > >> >> > Best wishes, > >> >> > > >> >> > Shinichi Nakagawa > >> >> > > >> >> > [[alternative HTML version deleted]] > >> >> > > >> >> > _______________________________________________ > >> >> > R-sig-phylo mailing list - R-sig-phylo@r-project.org > >> <mailto:R-sig-phylo@r-project.org> > >> >> <mailto:R-sig-phylo@r-project.org > >> <mailto:R-sig-phylo@r-project.org>> > >> >> > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > >> <https://stat.ethz.ch/mailman/listinfo/r-sig-phylo> > >> >> <https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > >> <https://stat.ethz.ch/mailman/listinfo/r-sig-phylo>> > >> >> > Searchable archive at > >> >> http://www.mail-archive.com/r-sig-phylo@r-project.org/ > >> <http://www.mail-archive.com/r-sig-phylo@r-project.org/> > >> >> <http://www.mail-archive.com/r-sig-phylo@r-project.org/ > >> <http://www.mail-archive.com/r-sig-phylo@r-project.org/>> > >> >> > > >> >> > > >> >> > Pour nous remonter une erreur de filtrage, veuillez vous > rendre > >> >> ici : http://f.security-mail.net/301dITcFgNy > >> <http://f.security-mail.net/301dITcFgNy> > >> >> <http://f.security-mail.net/301dITcFgNy > >> <http://f.security-mail.net/301dITcFgNy>> > >> >> > > >> >> > > >> >> > >> >> > >> >> > >> >> > >> > > >> > _______________________________________________ > >> > R-sig-phylo mailing list - R-sig-phylo@r-project.org > >> <mailto:R-sig-phylo@r-project.org> > >> > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > >> <https://stat.ethz.ch/mailman/listinfo/r-sig-phylo> > >> > Searchable archive at > >> http://www.mail-archive.com/r-sig-phylo@r-project.org/ > >> <http://www.mail-archive.com/r-sig-phylo@r-project.org/> > >> > >> _______________________________________________ > >> R-sig-phylo mailing list - R-sig-phylo@r-project.org > >> <mailto:R-sig-phylo@r-project.org> > >> https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > >> <https://stat.ethz.ch/mailman/listinfo/r-sig-phylo> > >> Searchable archive at > >> http://www.mail-archive.com/r-sig-phylo@r-project.org/ > >> <http://www.mail-archive.com/r-sig-phylo@r-project.org/> > >> > >> > >> > >> > > > [[alternative HTML version deleted]] > > _______________________________________________ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r- > sig-ph...@r-project.org/ > > [[alternative HTML version deleted]] > > > _______________________________________________ > R-sig-phylo mailing list - R-sig-phylo@r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-phylo > Searchable archive at http://www.mail-archive.com/r- > sig-ph...@r-project.org/ > -- Klaus Schliep Postdoctoral Fellow Revell Lab, University of Massachusetts Boston http://www.phangorn.org/ [[alternative HTML version deleted]] _______________________________________________ R-sig-phylo mailing list - R-sig-phylo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/r-sig-phylo@r-project.org/